projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nilfs2: apply readahead for recovery on mount
[linux-2.6-block.git] / fs / nilfs2 / super.c
CommitLineData
783f6184
RK		 1/*
2 * super.c - NILFS module and super block management.
3 *
4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 *
20 * Written by Ryusuke Konishi <ryusuke@osrg.net>
21 */
22/*
23 * linux/fs/ext2/super.c
24 *
25 * Copyright (C) 1992, 1993, 1994, 1995
26 * Remy Card (card@masi.ibp.fr)
27 * Laboratoire MASI - Institut Blaise Pascal
28 * Universite Pierre et Marie Curie (Paris VI)
29 *
30 * from
31 *
32 * linux/fs/minix/inode.c
33 *
34 * Copyright (C) 1991, 1992 Linus Torvalds
35 *
36 * Big-endian to little-endian byte-swapping/bitmaps by
37 * David S. Miller (davem@caip.rutgers.edu), 1995
38 */
39
40#include <linux/module.h>
41#include <linux/string.h>
42#include <linux/slab.h>
43#include <linux/init.h>
44#include <linux/blkdev.h>
45#include <linux/parser.h>
46#include <linux/random.h>
47#include <linux/crc32.h>
48#include <linux/smp_lock.h>
49#include <linux/vfs.h>
50#include <linux/writeback.h>
51#include <linux/kobject.h>
52#include <linux/exportfs.h>
b58a285b
JS		 53#include <linux/seq_file.h>
54#include <linux/mount.h>
783f6184
RK		 55#include "nilfs.h"
56#include "mdt.h"
57#include "alloc.h"
58#include "page.h"
59#include "cpfile.h"
60#include "ifile.h"
61#include "dat.h"
62#include "segment.h"
63#include "segbuf.h"
64
65MODULE_AUTHOR("NTT Corp.");
66MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
67 "(NILFS)");
783f6184
RK		 68MODULE_LICENSE("GPL");
69
70static int nilfs_remount(struct super_block *sb, int *flags, char *data);
783f6184
RK		 71
72/**
73 * nilfs_error() - report failure condition on a filesystem
74 *
75 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
76 * reporting an error message. It should be called when NILFS detects
77 * incoherences or defects of meta data on disk. As for sustainable
78 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
79 * function should be used instead.
80 *
81 * The segment constructor must not call this function because it can
82 * kill itself.
83 */
84void nilfs_error(struct super_block *sb, const char *function,
85 const char *fmt, ...)
86{
87 struct nilfs_sb_info *sbi = NILFS_SB(sb);
88 va_list args;
89
90 va_start(args, fmt);
91 printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function);
92 vprintk(fmt, args);
93 printk("\n");
94 va_end(args);
95
96 if (!(sb->s_flags & MS_RDONLY)) {
97 struct the_nilfs *nilfs = sbi->s_nilfs;
98
99 if (!nilfs_test_opt(sbi, ERRORS_CONT))
100 nilfs_detach_segment_constructor(sbi);
101
102 down_write(&nilfs->ns_sem);
103 if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
104 nilfs->ns_mount_state |= NILFS_ERROR_FS;
e339ad31
RK		 105 nilfs->ns_sbp[0]->s_state |=
106 cpu_to_le16(NILFS_ERROR_FS);
107 nilfs_commit_super(sbi, 1);
783f6184
RK		 108 }
109 up_write(&nilfs->ns_sem);
110
111 if (nilfs_test_opt(sbi, ERRORS_RO)) {
112 printk(KERN_CRIT "Remounting filesystem read-only\n");
113 sb->s_flags |= MS_RDONLY;
114 }
115 }
116
117 if (nilfs_test_opt(sbi, ERRORS_PANIC))
118 panic("NILFS (device %s): panic forced after error\n",
119 sb->s_id);
120}
121
122void nilfs_warning(struct super_block *sb, const char *function,
123 const char *fmt, ...)
124{
125 va_list args;
126
127 va_start(args, fmt);
128 printk(KERN_WARNING "NILFS warning (device %s): %s: ",
129 sb->s_id, function);
130 vprintk(fmt, args);
131 printk("\n");
132 va_end(args);
133}
134
135static struct kmem_cache *nilfs_inode_cachep;
136
a53b4751 137struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs)
783f6184
RK		 138{
139 struct nilfs_inode_info *ii;
140
141 ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS);
142 if (!ii)
143 return NULL;
144 ii->i_bh = NULL;
145 ii->i_state = 0;
146 ii->vfs_inode.i_version = 1;
a53b4751 147 nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi);
783f6184
RK		 148 return &ii->vfs_inode;
149}
150
a53b4751
RK		 151struct inode *nilfs_alloc_inode(struct super_block *sb)
152{
153 return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
154}
155
783f6184
RK		 156void nilfs_destroy_inode(struct inode *inode)
157{
158 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
159}
160
161static void init_once(void *obj)
162{
163 struct nilfs_inode_info *ii = obj;
164
165 INIT_LIST_HEAD(&ii->i_dirty);
166#ifdef CONFIG_NILFS_XATTR
167 init_rwsem(&ii->xattr_sem);
168#endif
169 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
170 ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union;
171 inode_init_once(&ii->vfs_inode);
172}
173
174static int nilfs_init_inode_cache(void)
175{
176 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
177 sizeof(struct nilfs_inode_info),
178 0, SLAB_RECLAIM_ACCOUNT,
179 init_once);
180
181 return (nilfs_inode_cachep == NULL) ? -ENOMEM : 0;
182}
183
184static inline void nilfs_destroy_inode_cache(void)
185{
186 kmem_cache_destroy(nilfs_inode_cachep);
187}
188
189static void nilfs_clear_inode(struct inode *inode)
190{
191 struct nilfs_inode_info *ii = NILFS_I(inode);
783f6184 192
783f6184
RK		 193 /*
194 * Free resources allocated in nilfs_read_inode(), here.
195 */
a2e7d2df 196 BUG_ON(!list_empty(&ii->i_dirty));
783f6184
RK		 197 brelse(ii->i_bh);
198 ii->i_bh = NULL;
783f6184
RK		 199
200 if (test_bit(NILFS_I_BMAP, &ii->i_state))
201 nilfs_bmap_clear(ii->i_bmap);
202
203 nilfs_btnode_cache_clear(&ii->i_btnode_cache);
783f6184
RK		 204}
205
e339ad31 206static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb)
783f6184
RK		 207{
208 struct the_nilfs *nilfs = sbi->s_nilfs;
209 int err;
210 int barrier_done = 0;
211
212 if (nilfs_test_opt(sbi, BARRIER)) {
e339ad31 213 set_buffer_ordered(nilfs->ns_sbh[0]);
783f6184
RK		 214 barrier_done = 1;
215 }
216 retry:
e339ad31
RK		 217 set_buffer_dirty(nilfs->ns_sbh[0]);
218 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
783f6184
RK		 219 if (err == -EOPNOTSUPP && barrier_done) {
220 nilfs_warning(sbi->s_super, __func__,
221 "barrier-based sync failed. "
222 "disabling barriers\n");
223 nilfs_clear_opt(sbi, BARRIER);
224 barrier_done = 0;
e339ad31 225 clear_buffer_ordered(nilfs->ns_sbh[0]);
783f6184
RK		 226 goto retry;
227 }
e339ad31 228 if (unlikely(err)) {
783f6184
RK		 229 printk(KERN_ERR
230 "NILFS: unable to write superblock (err=%d)\n", err);
e339ad31
RK		 231 if (err == -EIO && nilfs->ns_sbh[1]) {
232 nilfs_fall_back_super_block(nilfs);
233 goto retry;
234 }
235 } else {
236 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
237
238 /*
239 * The latest segment becomes trailable from the position
240 * written in superblock.
241 */
783f6184 242 clear_nilfs_discontinued(nilfs);
e339ad31
RK		 243
244 /* update GC protection for recent segments */
245 if (nilfs->ns_sbh[1]) {
246 sbp = NULL;
247 if (dupsb) {
248 set_buffer_dirty(nilfs->ns_sbh[1]);
249 if (!sync_dirty_buffer(nilfs->ns_sbh[1]))
250 sbp = nilfs->ns_sbp[1];
251 }
252 }
253 if (sbp) {
254 spin_lock(&nilfs->ns_last_segment_lock);
255 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
256 spin_unlock(&nilfs->ns_last_segment_lock);
257 }
783f6184
RK		 258 }
259
260 return err;
261}
262
e339ad31 263int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
783f6184
RK		 264{
265 struct the_nilfs *nilfs = sbi->s_nilfs;
e339ad31 266 struct nilfs_super_block **sbp = nilfs->ns_sbp;
783f6184 267 sector_t nfreeblocks;
e339ad31 268 time_t t;
783f6184
RK		 269 int err;
270
271 /* nilfs->sem must be locked by the caller. */
e339ad31
RK		 272 if (sbp[0]->s_magic != NILFS_SUPER_MAGIC) {
273 if (sbp[1] && sbp[1]->s_magic == NILFS_SUPER_MAGIC)
274 nilfs_swap_super_block(nilfs);
275 else {
276 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
277 sbi->s_super->s_id);
278 return -EIO;
279 }
280 }
783f6184
RK		 281 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
282 if (unlikely(err)) {
283 printk(KERN_ERR "NILFS: failed to count free blocks\n");
284 return err;
285 }
e339ad31
RK		 286 spin_lock(&nilfs->ns_last_segment_lock);
287 sbp[0]->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
288 sbp[0]->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
289 sbp[0]->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
290 spin_unlock(&nilfs->ns_last_segment_lock);
291
292 t = get_seconds();
293 nilfs->ns_sbwtime[0] = t;
294 sbp[0]->s_free_blocks_count = cpu_to_le64(nfreeblocks);
295 sbp[0]->s_wtime = cpu_to_le64(t);
296 sbp[0]->s_sum = 0;
297 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
298 (unsigned char *)sbp[0],
299 nilfs->ns_sbsize));
300 if (dupsb && sbp[1]) {
301 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
302 nilfs->ns_sbwtime[1] = t;
303 }
783f6184 304 sbi->s_super->s_dirt = 0;
e339ad31 305 return nilfs_sync_super(sbi, dupsb);
783f6184
RK		 306}
307
308static void nilfs_put_super(struct super_block *sb)
309{
310 struct nilfs_sb_info *sbi = NILFS_SB(sb);
311 struct the_nilfs *nilfs = sbi->s_nilfs;
312
6cfd0148
CH		 313 lock_kernel();
314
783f6184
RK		 315 nilfs_detach_segment_constructor(sbi);
316
317 if (!(sb->s_flags & MS_RDONLY)) {
318 down_write(&nilfs->ns_sem);
e339ad31
RK		 319 nilfs->ns_sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state);
320 nilfs_commit_super(sbi, 1);
783f6184
RK		 321 up_write(&nilfs->ns_sem);
322 }
e59399d0 323 down_write(&nilfs->ns_super_sem);
3f82ff55
RK		 324 if (nilfs->ns_current == sbi)
325 nilfs->ns_current = NULL;
e59399d0 326 up_write(&nilfs->ns_super_sem);
783f6184
RK		 327
328 nilfs_detach_checkpoint(sbi);
329 put_nilfs(sbi->s_nilfs);
330 sbi->s_super = NULL;
331 sb->s_fs_info = NULL;
6dd47406 332 nilfs_put_sbinfo(sbi);
6cfd0148
CH		 333
334 unlock_kernel();
783f6184
RK		 335}
336
783f6184
RK		 337static int nilfs_sync_fs(struct super_block *sb, int wait)
338{
6233caa9
JS		 339 struct nilfs_sb_info *sbi = NILFS_SB(sb);
340 struct the_nilfs *nilfs = sbi->s_nilfs;
783f6184
RK		 341 int err = 0;
342
343 /* This function is called when super block should be written back */
344 if (wait)
345 err = nilfs_construct_segment(sb);
6233caa9
JS		 346
347 down_write(&nilfs->ns_sem);
348 if (sb->s_dirt)
349 nilfs_commit_super(sbi, 1);
350 up_write(&nilfs->ns_sem);
351
783f6184
RK		 352 return err;
353}
354
355int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
356{
357 struct the_nilfs *nilfs = sbi->s_nilfs;
358 struct nilfs_checkpoint *raw_cp;
359 struct buffer_head *bh_cp;
360 int err;
361
e59399d0 362 down_write(&nilfs->ns_super_sem);
783f6184 363 list_add(&sbi->s_list, &nilfs->ns_supers);
e59399d0 364 up_write(&nilfs->ns_super_sem);
783f6184 365
79739565 366 sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size);
783f6184
RK		 367 if (!sbi->s_ifile)
368 return -ENOMEM;
369
1154ecbd 370 down_read(&nilfs->ns_segctor_sem);
783f6184
RK		 371 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
372 &bh_cp);
1154ecbd 373 up_read(&nilfs->ns_segctor_sem);
783f6184
RK		 374 if (unlikely(err)) {
375 if (err == -ENOENT || err == -EINVAL) {
376 printk(KERN_ERR
377 "NILFS: Invalid checkpoint "
378 "(checkpoint number=%llu)\n",
379 (unsigned long long)cno);
380 err = -EINVAL;
381 }
382 goto failed;
383 }
384 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
385 if (unlikely(err))
386 goto failed_bh;
387 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
388 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
389
390 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
391 return 0;
392
393 failed_bh:
394 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
395 failed:
396 nilfs_mdt_destroy(sbi->s_ifile);
397 sbi->s_ifile = NULL;
398
e59399d0 399 down_write(&nilfs->ns_super_sem);
783f6184 400 list_del_init(&sbi->s_list);
e59399d0 401 up_write(&nilfs->ns_super_sem);
783f6184
RK		 402
403 return err;
404}
405
406void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
407{
408 struct the_nilfs *nilfs = sbi->s_nilfs;
409
783f6184
RK		 410 nilfs_mdt_destroy(sbi->s_ifile);
411 sbi->s_ifile = NULL;
e59399d0 412 down_write(&nilfs->ns_super_sem);
783f6184 413 list_del_init(&sbi->s_list);
e59399d0 414 up_write(&nilfs->ns_super_sem);
783f6184
RK		 415}
416
417static int nilfs_mark_recovery_complete(struct nilfs_sb_info *sbi)
418{
419 struct the_nilfs *nilfs = sbi->s_nilfs;
420 int err = 0;
421
422 down_write(&nilfs->ns_sem);
423 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
424 nilfs->ns_mount_state |= NILFS_VALID_FS;
e339ad31 425 err = nilfs_commit_super(sbi, 1);
783f6184
RK		 426 if (likely(!err))
427 printk(KERN_INFO "NILFS: recovery complete.\n");
428 }
429 up_write(&nilfs->ns_sem);
430 return err;
431}
432
433static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
434{
435 struct super_block *sb = dentry->d_sb;
436 struct nilfs_sb_info *sbi = NILFS_SB(sb);
c306af23
RK		 437 struct the_nilfs *nilfs = sbi->s_nilfs;
438 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
783f6184
RK		 439 unsigned long long blocks;
440 unsigned long overhead;
441 unsigned long nrsvblocks;
442 sector_t nfreeblocks;
783f6184
RK		 443 int err;
444
445 /*
446 * Compute all of the segment blocks
447 *
448 * The blocks before first segment and after last segment
449 * are excluded.
450 */
451 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
452 - nilfs->ns_first_data_block;
453 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
454
455 /*
456 * Compute the overhead
457 *
458 * When distributing meta data blocks outside semgent structure,
459 * We must count them as the overhead.
460 */
461 overhead = 0;
462
463 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
464 if (unlikely(err))
465 return err;
466
467 buf->f_type = NILFS_SUPER_MAGIC;
468 buf->f_bsize = sb->s_blocksize;
469 buf->f_blocks = blocks - overhead;
470 buf->f_bfree = nfreeblocks;
471 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
472 (buf->f_bfree - nrsvblocks) : 0;
473 buf->f_files = atomic_read(&sbi->s_inodes_count);
474 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
475 buf->f_namelen = NILFS_NAME_LEN;
c306af23
RK		 476 buf->f_fsid.val[0] = (u32)id;
477 buf->f_fsid.val[1] = (u32)(id >> 32);
478
783f6184
RK		 479 return 0;
480}
481
b58a285b
JS		 482static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
483{
484 struct super_block *sb = vfs->mnt_sb;
485 struct nilfs_sb_info *sbi = NILFS_SB(sb);
486
487 if (!nilfs_test_opt(sbi, BARRIER))
91f1953b 488 seq_printf(seq, ",nobarrier");
b58a285b
JS		 489 if (nilfs_test_opt(sbi, SNAPSHOT))
490 seq_printf(seq, ",cp=%llu",
491 (unsigned long long int)sbi->s_snapshot_cno);
492 if (nilfs_test_opt(sbi, ERRORS_RO))
493 seq_printf(seq, ",errors=remount-ro");
494 if (nilfs_test_opt(sbi, ERRORS_PANIC))
495 seq_printf(seq, ",errors=panic");
496 if (nilfs_test_opt(sbi, STRICT_ORDER))
497 seq_printf(seq, ",order=strict");
498
499 return 0;
500}
501
b87221de 502static const struct super_operations nilfs_sops = {
783f6184
RK		 503 .alloc_inode = nilfs_alloc_inode,
504 .destroy_inode = nilfs_destroy_inode,
505 .dirty_inode = nilfs_dirty_inode,
506 /* .write_inode = nilfs_write_inode, */
507 /* .put_inode = nilfs_put_inode, */
508 /* .drop_inode = nilfs_drop_inode, */
509 .delete_inode = nilfs_delete_inode,
510 .put_super = nilfs_put_super,
1dfa2710 511 /* .write_super = nilfs_write_super, */
783f6184
RK		 512 .sync_fs = nilfs_sync_fs,
513 /* .write_super_lockfs */
514 /* .unlockfs */
515 .statfs = nilfs_statfs,
516 .remount_fs = nilfs_remount,
517 .clear_inode = nilfs_clear_inode,
518 /* .umount_begin */
b58a285b 519 .show_options = nilfs_show_options
783f6184
RK		 520};
521
522static struct inode *
523nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
524{
525 struct inode *inode;
526
527 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
528 ino != NILFS_SKETCH_INO)
529 return ERR_PTR(-ESTALE);
530
531 inode = nilfs_iget(sb, ino);
532 if (IS_ERR(inode))
533 return ERR_CAST(inode);
534 if (generation && inode->i_generation != generation) {
535 iput(inode);
536 return ERR_PTR(-ESTALE);
537 }
538
539 return inode;
540}
541
542static struct dentry *
543nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
544 int fh_type)
545{
546 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
547 nilfs_nfs_get_inode);
548}
549
550static struct dentry *
551nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
552 int fh_type)
553{
554 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
555 nilfs_nfs_get_inode);
556}
557
ac4cfdd6 558static const struct export_operations nilfs_export_ops = {
783f6184
RK		 559 .fh_to_dentry = nilfs_fh_to_dentry,
560 .fh_to_parent = nilfs_fh_to_parent,
561 .get_parent = nilfs_get_parent,
562};
563
564enum {
565 Opt_err_cont, Opt_err_panic, Opt_err_ro,
91f1953b 566 Opt_nobarrier, Opt_snapshot, Opt_order,
783f6184
RK		 567 Opt_err,
568};
569
570static match_table_t tokens = {
571 {Opt_err_cont, "errors=continue"},
572 {Opt_err_panic, "errors=panic"},
573 {Opt_err_ro, "errors=remount-ro"},
91f1953b 574 {Opt_nobarrier, "nobarrier"},
783f6184
RK		 575 {Opt_snapshot, "cp=%u"},
576 {Opt_order, "order=%s"},
577 {Opt_err, NULL}
578};
579
783f6184
RK		 580static int parse_options(char *options, struct super_block *sb)
581{
582 struct nilfs_sb_info *sbi = NILFS_SB(sb);
583 char *p;
584 substring_t args[MAX_OPT_ARGS];
585 int option;
586
587 if (!options)
588 return 1;
589
590 while ((p = strsep(&options, ",")) != NULL) {
591 int token;
592 if (!*p)
593 continue;
594
595 token = match_token(p, tokens, args);
596 switch (token) {
91f1953b
JS		 597 case Opt_nobarrier:
598 nilfs_clear_opt(sbi, BARRIER);
783f6184
RK		 599 break;
600 case Opt_order:
601 if (strcmp(args[0].from, "relaxed") == 0)
602 /* Ordered data semantics */
603 nilfs_clear_opt(sbi, STRICT_ORDER);
604 else if (strcmp(args[0].from, "strict") == 0)
605 /* Strict in-order semantics */
606 nilfs_set_opt(sbi, STRICT_ORDER);
607 else
608 return 0;
609 break;
610 case Opt_err_panic:
611 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
612 break;
613 case Opt_err_ro:
614 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
615 break;
616 case Opt_err_cont:
617 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
618 break;
619 case Opt_snapshot:
620 if (match_int(&args[0], &option) || option <= 0)
621 return 0;
622 if (!(sb->s_flags & MS_RDONLY))
623 return 0;
624 sbi->s_snapshot_cno = option;
625 nilfs_set_opt(sbi, SNAPSHOT);
626 break;
627 default:
628 printk(KERN_ERR
629 "NILFS: Unrecognized mount option \"%s\"\n", p);
630 return 0;
631 }
632 }
633 return 1;
634}
635
636static inline void
637nilfs_set_default_options(struct nilfs_sb_info *sbi,
638 struct nilfs_super_block *sbp)
639{
640 sbi->s_mount_opt =
641 NILFS_MOUNT_ERRORS_CONT | NILFS_MOUNT_BARRIER;
642}
643
644static int nilfs_setup_super(struct nilfs_sb_info *sbi)
645{
646 struct the_nilfs *nilfs = sbi->s_nilfs;
e339ad31 647 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
783f6184
RK		 648 int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count);
649 int mnt_count = le16_to_cpu(sbp->s_mnt_count);
650
651 /* nilfs->sem must be locked by the caller. */
652 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
653 printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
654 } else if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
655 printk(KERN_WARNING
656 "NILFS warning: mounting fs with errors\n");
657#if 0
658 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
659 printk(KERN_WARNING
660 "NILFS warning: maximal mount count reached\n");
661#endif
662 }
663 if (!max_mnt_count)
664 sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
665
666 sbp->s_mnt_count = cpu_to_le16(mnt_count + 1);
667 sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS);
668 sbp->s_mtime = cpu_to_le64(get_seconds());
e339ad31 669 return nilfs_commit_super(sbi, 1);
783f6184
RK		 670}
671
e339ad31
RK		 672struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
673 u64 pos, int blocksize,
674 struct buffer_head **pbh)
783f6184 675{
e339ad31
RK		 676 unsigned long long sb_index = pos;
677 unsigned long offset;
783f6184 678
e339ad31 679 offset = do_div(sb_index, blocksize);
783f6184 680 *pbh = sb_bread(sb, sb_index);
e339ad31 681 if (!*pbh)
783f6184 682 return NULL;
783f6184
RK		 683 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
684}
685
783f6184
RK		 686int nilfs_store_magic_and_option(struct super_block *sb,
687 struct nilfs_super_block *sbp,
688 char *data)
689{
690 struct nilfs_sb_info *sbi = NILFS_SB(sb);
691
783f6184
RK		 692 sb->s_magic = le16_to_cpu(sbp->s_magic);
693
694 /* FS independent flags */
695#ifdef NILFS_ATIME_DISABLE
696 sb->s_flags |= MS_NOATIME;
697#endif
698
783f6184
RK		 699 nilfs_set_default_options(sbi, sbp);
700
701 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
702 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
703 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
704 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
705
e339ad31 706 return !parse_options(data, sb) ? -EINVAL : 0 ;
783f6184
RK		 707}
708
709/**
710 * nilfs_fill_super() - initialize a super block instance
711 * @sb: super_block
712 * @data: mount options
713 * @silent: silent mode flag
714 * @nilfs: the_nilfs struct
715 *
aa7dfb89 716 * This function is called exclusively by nilfs->ns_mount_mutex.
783f6184
RK		 717 * So, the recovery process is protected from other simultaneous mounts.
718 */
719static int
720nilfs_fill_super(struct super_block *sb, void *data, int silent,
721 struct the_nilfs *nilfs)
722{
723 struct nilfs_sb_info *sbi;
724 struct inode *root;
725 __u64 cno;
726 int err;
727
728 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
729 if (!sbi)
730 return -ENOMEM;
731
732 sb->s_fs_info = sbi;
733
734 get_nilfs(nilfs);
735 sbi->s_nilfs = nilfs;
736 sbi->s_super = sb;
6dd47406 737 atomic_set(&sbi->s_count, 1);
783f6184
RK		 738
739 err = init_nilfs(nilfs, sbi, (char *)data);
740 if (err)
741 goto failed_sbi;
742
743 spin_lock_init(&sbi->s_inode_lock);
744 INIT_LIST_HEAD(&sbi->s_dirty_files);
745 INIT_LIST_HEAD(&sbi->s_list);
746
747 /*
748 * Following initialization is overlapped because
749 * nilfs_sb_info structure has been cleared at the beginning.
750 * But we reserve them to keep our interest and make ready
751 * for the future change.
752 */
753 get_random_bytes(&sbi->s_next_generation,
754 sizeof(sbi->s_next_generation));
755 spin_lock_init(&sbi->s_next_gen_lock);
756
757 sb->s_op = &nilfs_sops;
758 sb->s_export_op = &nilfs_export_ops;
759 sb->s_root = NULL;
61239230 760 sb->s_time_gran = 1;
783f6184
RK		 761
762 if (!nilfs_loaded(nilfs)) {
763 err = load_nilfs(nilfs, sbi);
764 if (err)
765 goto failed_sbi;
766 }
767 cno = nilfs_last_cno(nilfs);
768
769 if (sb->s_flags & MS_RDONLY) {
770 if (nilfs_test_opt(sbi, SNAPSHOT)) {
43be0ec0 771 down_read(&nilfs->ns_segctor_sem);
1f5abe7e
RK		 772 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
773 sbi->s_snapshot_cno);
43be0ec0
ZY		 774 up_read(&nilfs->ns_segctor_sem);
775 if (err < 0) {
776 if (err == -ENOENT)
777 err = -EINVAL;
1f5abe7e 778 goto failed_sbi;
43be0ec0 779 }
1f5abe7e 780 if (!err) {
783f6184
RK		 781 printk(KERN_ERR
782 "NILFS: The specified checkpoint is "
783 "not a snapshot "
784 "(checkpoint number=%llu).\n",
785 (unsigned long long)sbi->s_snapshot_cno);
786 err = -EINVAL;
787 goto failed_sbi;
788 }
789 cno = sbi->s_snapshot_cno;
790 } else
791 /* Read-only mount */
792 sbi->s_snapshot_cno = cno;
793 }
794
795 err = nilfs_attach_checkpoint(sbi, cno);
796 if (err) {
797 printk(KERN_ERR "NILFS: error loading a checkpoint"
798 " (checkpoint number=%llu).\n", (unsigned long long)cno);
799 goto failed_sbi;
800 }
801
802 if (!(sb->s_flags & MS_RDONLY)) {
cece5520 803 err = nilfs_attach_segment_constructor(sbi);
783f6184
RK		 804 if (err)
805 goto failed_checkpoint;
806 }
807
808 root = nilfs_iget(sb, NILFS_ROOT_INO);
809 if (IS_ERR(root)) {
810 printk(KERN_ERR "NILFS: get root inode failed\n");
811 err = PTR_ERR(root);
812 goto failed_segctor;
813 }
814 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
815 iput(root);
816 printk(KERN_ERR "NILFS: corrupt root inode.\n");
817 err = -EINVAL;
818 goto failed_segctor;
819 }
820 sb->s_root = d_alloc_root(root);
821 if (!sb->s_root) {
822 iput(root);
823 printk(KERN_ERR "NILFS: get root dentry failed\n");
824 err = -ENOMEM;
825 goto failed_segctor;
826 }
827
828 if (!(sb->s_flags & MS_RDONLY)) {
829 down_write(&nilfs->ns_sem);
830 nilfs_setup_super(sbi);
831 up_write(&nilfs->ns_sem);
832 }
833
834 err = nilfs_mark_recovery_complete(sbi);
835 if (unlikely(err)) {
836 printk(KERN_ERR "NILFS: recovery failed.\n");
837 goto failed_root;
838 }
839
e59399d0 840 down_write(&nilfs->ns_super_sem);
3f82ff55
RK		 841 if (!nilfs_test_opt(sbi, SNAPSHOT))
842 nilfs->ns_current = sbi;
e59399d0 843 up_write(&nilfs->ns_super_sem);
3f82ff55 844
783f6184
RK		 845 return 0;
846
847 failed_root:
848 dput(sb->s_root);
849 sb->s_root = NULL;
850
851 failed_segctor:
852 nilfs_detach_segment_constructor(sbi);
853
854 failed_checkpoint:
855 nilfs_detach_checkpoint(sbi);
856
857 failed_sbi:
858 put_nilfs(nilfs);
859 sb->s_fs_info = NULL;
6dd47406 860 nilfs_put_sbinfo(sbi);
783f6184
RK		 861 return err;
862}
863
864static int nilfs_remount(struct super_block *sb, int *flags, char *data)
865{
866 struct nilfs_sb_info *sbi = NILFS_SB(sb);
867 struct nilfs_super_block *sbp;
868 struct the_nilfs *nilfs = sbi->s_nilfs;
869 unsigned long old_sb_flags;
870 struct nilfs_mount_options old_opts;
871 int err;
872
337eb00a
AIB		 873 lock_kernel();
874
e59399d0 875 down_write(&nilfs->ns_super_sem);
783f6184
RK		 876 old_sb_flags = sb->s_flags;
877 old_opts.mount_opt = sbi->s_mount_opt;
878 old_opts.snapshot_cno = sbi->s_snapshot_cno;
879
880 if (!parse_options(data, sb)) {
881 err = -EINVAL;
882 goto restore_opts;
883 }
884 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
885
886 if ((*flags & MS_RDONLY) &&
887 sbi->s_snapshot_cno != old_opts.snapshot_cno) {
888 printk(KERN_WARNING "NILFS (device %s): couldn't "
889 "remount to a different snapshot. \n",
890 sb->s_id);
891 err = -EINVAL;
892 goto restore_opts;
893 }
894
895 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
896 goto out;
897 if (*flags & MS_RDONLY) {
898 /* Shutting down the segment constructor */
899 nilfs_detach_segment_constructor(sbi);
900 sb->s_flags |= MS_RDONLY;
901
902 sbi->s_snapshot_cno = nilfs_last_cno(nilfs);
903 /* nilfs_set_opt(sbi, SNAPSHOT); */
904
905 /*
906 * Remounting a valid RW partition RDONLY, so set
907 * the RDONLY flag and then mark the partition as valid again.
908 */
909 down_write(&nilfs->ns_sem);
e339ad31 910 sbp = nilfs->ns_sbp[0];
783f6184
RK		 911 if (!(sbp->s_state & le16_to_cpu(NILFS_VALID_FS)) &&
912 (nilfs->ns_mount_state & NILFS_VALID_FS))
913 sbp->s_state = cpu_to_le16(nilfs->ns_mount_state);
914 sbp->s_mtime = cpu_to_le64(get_seconds());
e339ad31 915 nilfs_commit_super(sbi, 1);
783f6184
RK		 916 up_write(&nilfs->ns_sem);
917 } else {
918 /*
919 * Mounting a RDONLY partition read-write, so reread and
920 * store the current valid flag. (It may have been changed
921 * by fsck since we originally mounted the partition.)
922 */
3f82ff55 923 if (nilfs->ns_current && nilfs->ns_current != sbi) {
783f6184 924 printk(KERN_WARNING "NILFS (device %s): couldn't "
3f82ff55 925 "remount because an RW-mount exists.\n",
783f6184
RK		 926 sb->s_id);
927 err = -EBUSY;
e59399d0 928 goto restore_opts;
783f6184
RK		 929 }
930 if (sbi->s_snapshot_cno != nilfs_last_cno(nilfs)) {
931 printk(KERN_WARNING "NILFS (device %s): couldn't "
932 "remount because the current RO-mount is not "
933 "the latest one.\n",
934 sb->s_id);
935 err = -EINVAL;
e59399d0 936 goto restore_opts;
783f6184
RK		 937 }
938 sb->s_flags &= ~MS_RDONLY;
939 nilfs_clear_opt(sbi, SNAPSHOT);
940 sbi->s_snapshot_cno = 0;
941
cece5520 942 err = nilfs_attach_segment_constructor(sbi);
783f6184 943 if (err)
e59399d0 944 goto restore_opts;
783f6184
RK		 945
946 down_write(&nilfs->ns_sem);
947 nilfs_setup_super(sbi);
948 up_write(&nilfs->ns_sem);
949
e59399d0 950 nilfs->ns_current = sbi;
783f6184
RK		 951 }
952 out:
e59399d0 953 up_write(&nilfs->ns_super_sem);
337eb00a 954 unlock_kernel();
783f6184
RK		 955 return 0;
956
783f6184
RK		 957 restore_opts:
958 sb->s_flags = old_sb_flags;
959 sbi->s_mount_opt = old_opts.mount_opt;
960 sbi->s_snapshot_cno = old_opts.snapshot_cno;
e59399d0 961 up_write(&nilfs->ns_super_sem);
337eb00a 962 unlock_kernel();
783f6184
RK		 963 return err;
964}
965
966struct nilfs_super_data {
967 struct block_device *bdev;
6dd47406 968 struct nilfs_sb_info *sbi;
783f6184
RK		 969 __u64 cno;
970 int flags;
971};
972
973/**
974 * nilfs_identify - pre-read mount options needed to identify mount instance
975 * @data: mount options
976 * @sd: nilfs_super_data
977 */
978static int nilfs_identify(char *data, struct nilfs_super_data *sd)
979{
980 char *p, *options = data;
981 substring_t args[MAX_OPT_ARGS];
982 int option, token;
983 int ret = 0;
984
985 do {
986 p = strsep(&options, ",");
987 if (p != NULL && *p) {
988 token = match_token(p, tokens, args);
989 if (token == Opt_snapshot) {
990 if (!(sd->flags & MS_RDONLY))
991 ret++;
992 else {
993 ret = match_int(&args[0], &option);
994 if (!ret) {
995 if (option > 0)
996 sd->cno = option;
997 else
998 ret++;
999 }
1000 }
1001 }
1002 if (ret)
1003 printk(KERN_ERR
1004 "NILFS: invalid mount option: %s\n", p);
1005 }
1006 if (!options)
1007 break;
1008 BUG_ON(options == data);
1009 *(options - 1) = ',';
1010 } while (!ret);
1011 return ret;
1012}
1013
1014static int nilfs_set_bdev_super(struct super_block *s, void *data)
1015{
1016 struct nilfs_super_data *sd = data;
1017
1018 s->s_bdev = sd->bdev;
1019 s->s_dev = s->s_bdev->bd_dev;
1020 return 0;
1021}
1022
1023static int nilfs_test_bdev_super(struct super_block *s, void *data)
783f6184
RK		 1024{
1025 struct nilfs_super_data *sd = data;
6dd47406
RK		 1026
1027 return sd->sbi && s->s_fs_info == (void *)sd->sbi;
783f6184
RK		 1028}
1029
1030static int
1031nilfs_get_sb(struct file_system_type *fs_type, int flags,
1032 const char *dev_name, void *data, struct vfsmount *mnt)
1033{
1034 struct nilfs_super_data sd;
33c8e57c
RK		 1035 struct super_block *s;
1036 struct the_nilfs *nilfs;
783f6184
RK		 1037 int err, need_to_close = 1;
1038
1039 sd.bdev = open_bdev_exclusive(dev_name, flags, fs_type);
1040 if (IS_ERR(sd.bdev))
1041 return PTR_ERR(sd.bdev);
1042
1043 /*
1044 * To get mount instance using sget() vfs-routine, NILFS needs
1045 * much more information than normal filesystems to identify mount
1046 * instance. For snapshot mounts, not only a mount type (ro-mount
1047 * or rw-mount) but also a checkpoint number is required.
783f6184
RK		 1048 */
1049 sd.cno = 0;
1050 sd.flags = flags;
1051 if (nilfs_identify((char *)data, &sd)) {
1052 err = -EINVAL;
1053 goto failed;
1054 }
1055
33c8e57c
RK		 1056 nilfs = find_or_create_nilfs(sd.bdev);
1057 if (!nilfs) {
1058 err = -ENOMEM;
1059 goto failed;
1060 }
1061
aa7dfb89 1062 mutex_lock(&nilfs->ns_mount_mutex);
3f82ff55
RK		 1063
1064 if (!sd.cno) {
1065 /*
1066 * Check if an exclusive mount exists or not.
1067 * Snapshot mounts coexist with a current mount
1068 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1069 * ro-mount are mutually exclusive.
1070 */
e59399d0 1071 down_read(&nilfs->ns_super_sem);
3f82ff55
RK		 1072 if (nilfs->ns_current &&
1073 ((nilfs->ns_current->s_super->s_flags ^ flags)
1074 & MS_RDONLY)) {
e59399d0 1075 up_read(&nilfs->ns_super_sem);
3f82ff55
RK		 1076 err = -EBUSY;
1077 goto failed_unlock;
1078 }
e59399d0 1079 up_read(&nilfs->ns_super_sem);
783f6184
RK		 1080 }
1081
1082 /*
6dd47406 1083 * Find existing nilfs_sb_info struct
783f6184 1084 */
6dd47406
RK		 1085 sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
1086
6dd47406
RK		 1087 /*
1088 * Get super block instance holding the nilfs_sb_info struct.
1089 * A new instance is allocated if no existing mount is present or
1090 * existing instance has been unmounted.
1091 */
33c8e57c 1092 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
6dd47406
RK		 1093 if (sd.sbi)
1094 nilfs_put_sbinfo(sd.sbi);
1095
33c8e57c
RK		 1096 if (IS_ERR(s)) {
1097 err = PTR_ERR(s);
1098 goto failed_unlock;
783f6184
RK		 1099 }
1100
1101 if (!s->s_root) {
1102 char b[BDEVNAME_SIZE];
1103
33c8e57c 1104 /* New superblock instance created */
783f6184
RK		 1105 s->s_flags = flags;
1106 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1107 sb_set_blocksize(s, block_size(sd.bdev));
1108
1109 err = nilfs_fill_super(s, data, flags & MS_VERBOSE, nilfs);
1110 if (err)
1111 goto cancel_new;
1112
1113 s->s_flags |= MS_ACTIVE;
1114 need_to_close = 0;
783f6184
RK		 1115 }
1116
aa7dfb89 1117 mutex_unlock(&nilfs->ns_mount_mutex);
783f6184
RK		 1118 put_nilfs(nilfs);
1119 if (need_to_close)
1120 close_bdev_exclusive(sd.bdev, flags);
1121 simple_set_mnt(mnt, s);
1122 return 0;
1123
783f6184 1124 failed_unlock:
aa7dfb89 1125 mutex_unlock(&nilfs->ns_mount_mutex);
33c8e57c 1126 put_nilfs(nilfs);
783f6184
RK		 1127 failed:
1128 close_bdev_exclusive(sd.bdev, flags);
1129
1130 return err;
1131
1132 cancel_new:
1133 /* Abandoning the newly allocated superblock */
aa7dfb89 1134 mutex_unlock(&nilfs->ns_mount_mutex);
33c8e57c 1135 put_nilfs(nilfs);
783f6184
RK		 1136 up_write(&s->s_umount);
1137 deactivate_super(s);
1138 /*
1139 * deactivate_super() invokes close_bdev_exclusive().
1140 * We must finish all post-cleaning before this call;
aa7dfb89 1141 * put_nilfs() needs the block device.
783f6184
RK		 1142 */
1143 return err;
1144}
1145
783f6184
RK		 1146struct file_system_type nilfs_fs_type = {
1147 .owner = THIS_MODULE,
1148 .name = "nilfs2",
1149 .get_sb = nilfs_get_sb,
1150 .kill_sb = kill_block_super,
1151 .fs_flags = FS_REQUIRES_DEV,
1152};
1153
1154static int __init init_nilfs_fs(void)
1155{
1156 int err;
1157
1158 err = nilfs_init_inode_cache();
1159 if (err)
1160 goto failed;
1161
1162 err = nilfs_init_transaction_cache();
1163 if (err)
1164 goto failed_inode_cache;
1165
1166 err = nilfs_init_segbuf_cache();
1167 if (err)
1168 goto failed_transaction_cache;
1169
1170 err = nilfs_btree_path_cache_init();
1171 if (err)
1172 goto failed_segbuf_cache;
1173
1174 err = register_filesystem(&nilfs_fs_type);
1175 if (err)
1176 goto failed_btree_path_cache;
1177
1178 return 0;
1179
1180 failed_btree_path_cache:
1181 nilfs_btree_path_cache_destroy();
1182
1183 failed_segbuf_cache:
1184 nilfs_destroy_segbuf_cache();
1185
1186 failed_transaction_cache:
1187 nilfs_destroy_transaction_cache();
1188
1189 failed_inode_cache:
1190 nilfs_destroy_inode_cache();
1191
1192 failed:
1193 return err;
1194}
1195
1196static void __exit exit_nilfs_fs(void)
1197{
1198 nilfs_destroy_segbuf_cache();
1199 nilfs_destroy_transaction_cache();
1200 nilfs_destroy_inode_cache();
1201 nilfs_btree_path_cache_destroy();
1202 unregister_filesystem(&nilfs_fs_type);
1203}
1204
1205module_init(init_nilfs_fs)
1206module_exit(exit_nilfs_fs)
Linux 6.x block layer and io_uring tree(s)